Simple Lora for generating Odo from Deep Space Nine (Played by the late Rene Auberjonois).Now updated for SDXL! I'm still learning how to fine-tune SDXL but my first attempts are already even better than anything a 1.5 LoRA could do. It's a little bit overtrained, so I recommend a Lora strength of 0.8 or lower.Only tested on the base SDXL model. Do not use the SDXL refiner or it ruins the Odo likeness.

This is only my second attempt to train a version of the Odo model on SDXL. Dataset is slightly changed, including some more images of Odo in alternate outfits so it does slightly better at clothing other than his normal uniform. Trained in 4 epochs on a 3080 in ~25min, I'm pretty impressed with how easy SDXL fine-tuning really is once you get the right parameters to get it working at all.

Files with SHA256: 8186cb38e7019ee9ff88fb3fe9de6beec00f87533922d8bfdc5cbc1f3781ff98